1069
THE AREA AND THE ORIENTATION ANGLES OF LEAVES IN MAIZE
CANOPY FROM STEREOVISION
N Ivanov 1 p Boissard *(*), M Chapron P Valéry *
1 INRA, Bioclimatologie, 78850 Thiverval-Grigrton, France.
2 ENSEA, Equipe de Traitement d'images et du Signal, 95014 Cergy-Pontoise, France.
ABSTRACT
3-D models of vegetation are elaborated to allow various simulations such as : interception of incoming
radiation, radiative transfer, reflectance simulation in remote sensing, study of plant functioning and inter
species competition. The construction of such 3-D models, based upon simple geometrical assumptions and
several production rules (growing, branching,...) poses two questions : how to acquire a large amount of leaves'
heights and orientation angles data in the field, and how to validate the model and assess its representativeness
at the scale of the field.
This paper presents a method of 3-D acquisition and reconstruction based upon stereovision, i.e. on the
acquisition of data using two synchronized CCD or photographic cameras. It shows first that Toscani's method
(Toscani, 1988) of camera calibration can be used in field's real conditions. The method is then applied to a tall
maize canopy (2,50 m high) because this canopy allows easy control measurements in the field. A destructive
procedure was used to acquire the data level by level so that a full canopy model could be elaborated. The
method proved to be satisfactory to provide rather good values of leaf area and inclination angles. The
geometrical model of the crop allowed to estimate statistical descriptors such as leaf area as a function of height
and leaf angular distribution. Finally some perspectives are presented.
KEY-WORDS:
Zea mays L = maize / stereovision / 3-D structure / geometry of vegetation / leaf area index / leaf angle
distribution / 3D computer model / plant profile / radiative transfer
1 - INTRODUCTION
Radiative transfer simulation models inside the canopy require to enter structural parameters such as leaf area
index (LAI) per vertical layer and leaf inclination angle distribution. This type of models, such as the SAIL
model (Verhoef, 1984) or the SAIL model enhanced by the hot spot effect (Kuusk, 1991), are particularly
important in remote sensing since model inversion is a mean to retrieve the relationship "spectral
reflectance =f( target's characteristics )".
In general, the plant geometry is rather poorly described since uniform variables such as the leaf area
index are used. Recently, computer vegetation models which give a good representation of 3-D geometry (de
Refiye et al, 1988) have been developped to study the interception of incoming solar radiation, directional
reflectance simulation using ray tracing (Dauzat et Hautecoeur, 1991; Lewis and Muller, 1992) as well as plant
growing or interspecies competition. This latter type of model, based on a geometric and topologie architecture
modeling of plant populations, requires a great lot of in situ measurements, which is time consuming.
Computer vegetation 3-D models of maize, elaborated from mathematical rules or statistical distributions, had
been proposed (Goel et al, 1991; Prévôt et al, 1991). They are based on very numerous measurements of
lengths and inclinations angles in the field carried out by an operator.
Different methods have been proposed to assess global structure parameters which do not explicitly
refer to individual plant or leaf. However, in the same time, various attempts were made to measure parameters
related to the fine 3-D structure such as the shape and the relative position of phytoelements:
- 2-D plant profiles (Bonhomme and Varlet-Grancher, 1978);
- 3-D digitizer (Moulia and Sinoquet, 1993);
- manual photogrammetric measurements (Boissard, 1985; Lewis and Muller, 1990);
- phytoelements locating using radiance measurements (Andrieu and Ivanov, 1993).
